Personally, I can not imagine a naive 20 year old gal, who had just been told that her boyfriend was not backing up her alibi anymore, calling the police LIARS! So what the heck do you do if the police do not believe your story, calling you a liar time and time again?
What a nightmare that interrogation must have been for Miss Knox.
Too bad the police did not record it.
Hmmm...
RWVBWL
Too right, in my opinion.
I think many people underestimate the power of well-established and well-tested police techniques to elicit information out of suspects. The problem is, these techniques are sometimes SO powerful that they elicit what could be termed "false positives".
For a very long time, the very existence of such "false positives" was either concealed or flat-out denied within law enforcement circles. However, when enough such cases had been definitively discovered, the law enforcement position had no option but to change. But only to something along the lines of "Sometimes mistakes (false confessions/accusations) happen, but exceptionally rarely and in very certain circumstances that we've already identified and eliminated. Juries should still draw very strong and certain inferences from any and all confessions/accusations obtained under interview".
Unfortunately, this "modified" position is itself being called more and more into question. Many recent - and scientifically controlled - studies show how amazingly (in the literal sense) easy it is to obtain false confessions and/or accusations when a person is subjected to a sophisticated combination of verbal abuse, coercion, suggestion, perception of personal peril, "Hobson's choice" games, "prisoner's dilemma" games and reward.
I would therefore make this argument (of my personal opinion): Yes, the police ARE generally very good at getting useful information out of suspects, if those suspects are indeed culpable of the crime in question. And sometimes they get whole confessions using these sorts of techniques. But not always. And evidence increasingly supports the view that a significant number of confessions obtained using this battery of techniques subsequently turn out to be false.
And so I'd argue this important point, which involves revisiting Bayesian conditional probability theory: IF you (i.e. the police) have a confession/accusation in your hand, AND the confession was obtained not spontaneously but only after employing these sorts of techniques, THEN the probability that the suspect has made an accurate confession/accusation (given the existence of confession and the techniques employed) is a LOT lower than 100%. Exactly how low, I can't currently say. I would be happy to do some digging around to try to put a number to it - that might take a little time though. However, I'm confident that the kernel of my argument, and the general principles, are accurate.
I'd also add that I believe (can probably find supporting evidence) that when people "break" and confess under these conditions, they generally confess to the whole, unvarnished truth - in a mixture of cathartic relief and a hope that "telling the truth" will help them in future judicial rulings. AK demonstrably didn't do this - and it's hard to imagine how she'd have thought she was either helping herself or healing herself through what she said at around 1.45am on 6th November...